Principles for a successful computerized physician order entry implementation.

Principles for a Successful Computerized Physician Order Entry Implementation Joan S. Ash, Ph.D., Lara Fournier, M.S., P. Zoë Stavri, Ph.D., Richard Dykstra, M.D. Department of Medical Informatics and Clinical Epidemiology, School of Medicine Oregon Health & Science University, Portland, OR To identify success factors for implementing computerized physician order entry (CPOE), our research team took both a top-down and bottom-up approach and reconciled the results to develop twelve overarching principles to guide implementation. A consensus panel of experts produced ten Considerations with nearly 150 subconsiderations, and a three year project using qualitative methods at multiple successful sites for a grounded theory approach yielded ten general themes with 24 sub-themes. After reconciliation using a meta-matrix approach, twelve Principles, which cluster into groups forming the mnemonic CPOE emerged. Computer technology principles include: temporal concerns; technology and meeting information needs; multidimensional integration; and costs. Personal principles are: value to users and tradeoffs; essential people; and training and support. Organizational principles include: foundational underpinnings; collaborative project management; terms, concepts and connotations; and improvement through evaluation and learning. Finally, Environmental issues include the motivation and context for implementing such systems. INTRODUCTION Computerized physician order entry (CPOE) continues to receive attention both because it has been shown to decrease medical errors [1-5], and because it has often been met with resistance on the part of users [6-9]. The National Library of Medicine awarded a three-year research grant to the POE Team (POET) at Oregon Health & Science University to do a field study to identify success factors for implementing CPOE. This was accomplished using two distinctly different approaches: a top-down assessment generated through a consensus conference of thirteen experts held in the spring of 2001; and a bottom-up, grounded theory, approach combining ethnographic and interview methods in the field. While there was considerable overlap and duplication of ideas in these two large data sets, there were also differences. Although each of the two sets of results was useful in its own right, members of POET felt compelled to reconcile them, to establish trustworthiness (the qualitative analog to validation) and to gain new insight. METHODS The Top-Down Approach A two day meeting at the Menucha retreat center near Portland, Oregon involved invited experts from around the world representing multiple stakeholder groups: clinicians, social scientists, information technology implementers, and vendors. The format was designed to stimulate creative discussion and consensus building and included brainstorming and storytelling sessions as well as small group work. All sessions were audiotaped and transcribed. The data were analyzed qualitatively [10]. All statements from the official typed notes and transcripts were printed out on separate cards (about 500 major statements). Five researchers used a card sort technique to produce ten categories. The conference attendees called them ‘Considerations’ because they should be pondered by those thinking about imp lementing CPOE rather than interpreted as strict guidelines. A consensus statement was generated after several months of online discussion. The list of ten Considerations is given in Table 1. Details are available as a list [at cpoe.org] and in text form [11]. The Bottom-Up Approach Four hospitals with successful CPOE implementations (defined as having over 80% of orders entered this way) were selected for field study based on geography, type of hospital, and length of system use. Two basic methods were used: ethnographic observation and interviews. Observation by research team members with different backgrounds used a common frame of reference and focus. Interviews were primarily oral history interviews of administrators, clinicians, and technology staff. Several focus group interviews were held with house officers and other clinicians. Researchers typed their own fieldnotes; interview tapes were transcribed by experienced oral history transcriptionists. Data were entered into N5 (formerly QSR NUD*IST, Sage Publications) to assist analysis. The multidisciplinary research team members individually coded and analyzed the transcripts and fieldnotes, then met as a team to agree on overarching themes a total of 33 times. The ten AMIA 2003 Symposium Proceedings − Page 36 Themes are outlined in Table 2 and have been described in prior papers [12-14]. The Reconciliation A meta-matrix was developed with the ten Themes along one axis and the ten Considerations along the other, each with all sub-themes listed as well. Metamatrices are used to organize qualitative data visually for further analysis [15]. This method is especially helpful for analyses done by teams and to merge different qualitative data sets [16]. The metarelationships are exceedingly complex, and the data behind them are rich and often subtle. While condensing the vast amount of data into categories is useful to illuminate the main points, important details may be excluded. In our process, we identified rows and columns and then asked questions about each cell (is there overlap or not and what is the nature of the overlap?); team members were thus forced to think about the underlying data. When questions could not be answered from memory, the original data were reviewed using the qualitative analysis software. Three team members who knew the data well “voted” on the amount of overlap in each cell in the matrix. For example, one cell was at the intersection of the Consideration “motivation for implementing POE” and the Theme “context.” Since the motivation might include pressure from outside the hospital and since the context might involve pressure from outside the hospital, each researcher would consider this a strong overlap. There were some strong overlaps, some weak ones, and some areas without overlap. The entire team met to view the matrix with the voting indicated on it and identified strong differences so that they could be further explored. RESULTS Table 2 lists twelve areas we called Principles. They represent the combination of the ten Themes from the grounded data and the ten Considerations developed by the experts. Each one is somewhat different from any individual Consideration or Theme. Two Principles emanate from only one of the data sets, but are important and the lack of overlap needs explanation. “Costs” were a Consideration in the data set that resulted from the Menucha consensus conference of experts and did not appear in the data set based on fieldwork, probably because users are not very concerned with cost. “Terms, concepts, and connotations” constituted a Theme in the data set from the fieldwork but were not part of the Considerations outlin (...truncated)